Introduction

Land–ocean linkage supports high productivity in coastal ecosystems, highlighting the importance of terrestrial input of both nutrients and particulate organic matter (POM) into the sea (Sholkovitz 1976; Matsunaga et al. 1998; Tsuda et al. 2003). Unlike the case with these forms of allochthonous input, not much attention has been paid to coarse organic matter, represented by vascular plant debris, as food sources for marine organisms. This is mainly because coarse plant debris has been regarded as resistant to biological decay. Nevertheless, coarse plant debris can harbor a dense and diverse assemblage of organisms in the sea (i.e., the sunken wood community) (in deep-sea floor: Bienhold et al. 2013; in mangrove swamp: Laurent et al. 2013), and wood-derived organic matter serves as one of the main carbon sources for the community, especially in the deep sea (Nishimoto et al. 2009).

In temperate shallow waters, coarse woody debris is fragmented by aggregation of wood borers, such as species of the molluscan bivalve families Teredinidae and Pholadidae and species of the crustacean isopod family Limnoriidae. Sedentary teredinid bivalves intensively fragment wood inside the logs, and the resulting honeycombed structure is collapsed from the surface by the attack of free-living limnoriid isopods (Nishimoto et al. 2015). These wood borers accelerate bacterial and fungal decomposition of the woody constituents in detritus (reviewed in Cragg et al. 2020) by increasing the surface/volume ratio of wood logs (Camilleri and Ribi 1986; Arnosti et al. 2018). When these wood borers are eaten by, for example, fish or predatory polychaetes (Russell 1997), organic matter derived from wood logs flows into higher trophic levels in the ecosystem as a result of predator–prey interactions. Thus, wood borers control the availability of food resources to members of the biological community and accelerate the assimilation of woody components into marine ecosystems.

A number of studies have addressed the ability of marine wood borers, especially members of the bivalve family Teredinidae, also known as shipworms, to digest wood logs. The teredinids harbor endosymbiotic cellulose-digesting and nitrogen-fixing bacteria in their gills (Distel et al. 2002). The bacteria-encoded enzymes are translocated from the gills to the gut and support the digestion of wood cellulose (O’Connor et al. 2014). Pechenik et al. (1979), however, reported the ability of a teredinid, Lyrodus pedicellatus Quatrefages, 1849, to assimilate phytoplankton in a feeding experiment using 14C-labeled one, and some anatomical studies have supported this finding (Turner 1966; Lopes et al. 2000). POM, consisting of phytoplankton and detritus, has been considered a supplemental food source for teredinids, but the relative importance of wood log and POM in their diet is controversial (Turner 1966). For example, recent stable isotope techniques showed inconsistent results; a teredinid, Teredo navalis Linnaeus, 1758, mainly use POM as the carbon source (Paalvast and van der Velde 2013), whereas the main carbon source of another teredinid, Bankia carinata Gray, 1827, is wood log (Charles et al. 2018). The limnoriid isopods have the ability to digest lignocellulose without the help of symbiotic microbes (King et al. 2010), and their digestive tracts are always filled only with woody particles (e.g., Sleeter et al. 1978). However, their nitrogen source is still unknown, and it might serve as carbon source, too. In contrast, the obligate wood-boring pholadid bivalve, Martesia striata, Linnaeus, 1758, is believed to be a true filter feeder judging from its anatomical features (Purchon 1956; Turner and Johnson 1971; Haga and Kase 2011), but there have been no studies of its cellulase activity. Therefore, it remains a critical issue whether these marine wood borers depend on food sources in addition to wood during the process of fragmentation of wood logs.

Carbon and nitrogen stable isotope ratios (hereafter referred to as δ13C and δ15N, respectively) have often been used to estimate the diets of target species. Generally, the δ13C value of a target species is slightly higher (around 0–1‰) than that of its prey (Hobson and Welch 1992), and a target species becomes enriched in δ15N value relative to its prey by 3–4‰ (Peterson and Fry 1987). These differences are the so-called “trophic enrichment factors” (TEF). By using these TEF values and the δ13C and δ15N values of potential food sources, the simple linear stable isotope mixing model approach can uniquely determine the proportional contributions of up to three different food sources based on mass balance equations (Phillips 2001). However, wood logs are poor in nitrogen and provide little nitrogen to xylophages (wood eaters). Furthermore, symbiotic nitrogen-fixing microbes fix dissolved N2 gas, and the resulting compounds are taken up by teredinid bivalves (Waterbury et al. 1983; Lechene et al. 2007). N2 gas cannot be a carbon source for teredinid bivalves. Therefore, these two parameters should be used to trace carbon and nitrogen sources for wood borers independently. This use is sufficiently effective in the case with only two different food sources like carbon source of the teredinids, and provides useful information in other cases too.

TEF values are often quoted from past studies dealing with similar relationships, because it is not realistic to calculate TEF values for every prey–predator relationship. The quoted TEF values are helpful for rough estimation of the proportional contribution of each potential food source in complex food web structures. In xylophages, however, TEF values are hard to determine, because their potential food sources are mixtures of labile and refractory organic matter with different stable isotope ratios. For example, teredinid bivalves selectively assimilate the labile components from wood logs (i.e., cellulose and hemicellulose). Among them, teredinid bivalves mainly utilize cellulose (Sabbadin et al 2018), which has a δ13C value approximately 3‰ higher than that of lignin and 1–2‰ higher than that of wood as a whole (Loader et al. 2003, in one of the oak woods, Quercus robur L.). In previous studies, the δ13C of whole wood was generally used as one of the end members in the mixing model without considering the gaps in δ13C values between wood logs and wood cellulose. In addition, the gaps are not necessarily 1–2‰ because of the differences in lignin contents among tree species. These discrepancies might lead to misinterpretation of the proportional contribution of each potential food source for xylophages in the use of the stable isotope mixing model.

The δ13C value of wood cellulose is known to fluctuate, together with that of wood as a whole (Loader et al. 2003), and, therefore, the variation in δ13C value of wood as a whole can be regarded as that of cellulose. POM, another independent carbon source of teredinid bivalves, is mainly consisted with labile marine phytoplankton and unidentifiable detrital particle. The δ13C value of POM is also known to fluctuate seasonally (e.g., Vizzini and Mazzola 2003). This variation is attributed to the seasonal changes in growth speed of marine phytoplankton (Laws et al. 1995) and/or in intensity of riverine organic inputs. In the area far from large river mouth, the fluctuation in δ13C values of POM can be regarded as the fluctuation in δ13C value of marine phytoplankton. Indeed, δ13C value of POM in this study site was the highest in summer, supporting this assumption. Therefore, as an alternative approach to the mixing model, we propose a basic multiple linear regression model in which the δ13C values of wood logs and POM themselves are adopted as parameters. We traced the seasonal and environmental changes in δ13C values of teredinid bivalves cultured under the same experimental conditions and related these changes to the variations in δ13C values of potential food sources by using a multiple linear regression model. This approach is advantageous in that it can estimate the proportional contribution of each potential carbon sources without the use of uncertain TEF values.

Materials and methods

Sampling design

In September 2008, we placed pieces of logs (ca. 10 cm in diameter and ca. 20 cm in length) of Japanese cedar, Cryptomeria japonica (Thunb. ex L.f.) D. Don, at ca. 2 m depth on the muddy bottom of Tanabe Bay, Kii Peninsula, on the Pacific coast of central Honshu, Japan (33°40′46.80″ N, 135°21′46.00″ E). The logs were fixed to the top of a polypropylene net that had been anchored to the sea floor. We randomly collected three log samples every 2 months for the first 16 months, and every 4 months thereafter, for a total of 48 months. At each sampling month, we also sampled 1 L of seawater from the sea bottom. Details of the experimental settings are given in Nishimoto et al. (2015).

After collecting the organisms attached to the wood surface, we manually broke the logs into small pieces and collected the wood-boring invertebrates. After measuring the shell length of the wood-boring bivalves, we analyzed the stable isotope ratios in tissues of five teredinid species, Teredo navalis Linnaeus, 1758, Teredo bartschi Clapp, 1923, Teredo clappi Bartsch, 1923, Lyrodus pedicellatus, and Nototeredo edax, Hedley, 1895, which was once misidentified as Psiloteredo megotara, Hanley in Forbes & Hanley, 1848 in Nishimoto et al. (2015), one pholadid bivalve, Martesia striata, and two limnoriid isopods, Limnoria tuberculata Sowinsky, 1884 and Limnoria saseboensis Menzies, 1957. Four epibenthic bivalve species, Brachidontes mutabilis, Gould, 1861, Striarca symmetrica, Reeve, 1844, Barbatia virescens, Reeve, 1844 and Crassostrea gigas, Thunberg, 1793 were also analyzed as examples of non-wood-boring bivalves.

Preparation of samples for stable isotope analysis

Animal samples collected during 2008–2012 were kept at − 20 °C until 2012 and were thereafter preserved in 99.5% ethanol at room temperature until further processing in 2016–2017 for stable isotope analysis. Freezing is the most common method to preserve samples for stable isotope analysis, because exposure to storage solution such as ethanol and formalin results in the changes in stable isotope values. The changes in δ13C values caused by ethanol-preservation are mainly due to the loss of lipid contents (Post et al. 2007), whereas organic solvents, such as chloroform–methanol solutions (Bligh and Dyer 1959), are generally used for the degreasing protocol in stable isotope analysis (Elliot et al. 2017). This is because δ13C is depleted in lipids compared to proteins, and there are variations in lipid content among specimens depending on the body conditions (Elliot et al. 2017). This degreasing procedure is, therefore, essential to compare food sources between specimens in different body conditions. Besides this degreasing effect, the loss or alteration of proteins also might affect stable isotope values. However, the shifts in δ13C and δ15N values after ethanol preservation were less than 0.5 ‰ for both the muscle tissues of marine fishes (average + 0.34 ‰ for δ13C and + 0.45 ‰ for δ15N in Durante et al. 2020, based on the review of past studies) and the foot tissues of freshwater bivalves (Syväranta et al. 2010). Furthermore, the potential changes in δ13C and δ15N values were suppressed in the multiple linear regression model. This is because the multiple linear regression model deals with the variation in stable isotope ratios, not the values themselves. The variations would be maintained among muscle tissue under the same preanalytical procedures (Arrington and Winemiller 2002), i.e., sample storage methods (presence or absence of fixation, type of fixative solution, and length of treatment) and tissue excision methods.

In molluscan specimens, δ13C and δ15N values differ among tissues, such as muscle, gonad, gill, digestive diverticula, and shell (McConnaughey and Gillikin 2008). In this study, adductor muscles were selectively used for stable isotope analysis because δ13C values in muscle tissues are expected to be less affected by ethanol preservation as mentioned above. To obtain sufficient volume for stable isotope analysis, larger specimens were selected from the sample. If there were no large specimens, siphons and/or mantles were used for the analysis in addition to adductor muscles. In the smaller pholadid bivalve and the limnoriid isopod specimens, the whole tissues of several specimens, from which the digestive tracts were thoroughly removed, were mixed. Using large pholadid specimens, we also investigated the variations in δ13C and δ15N values among body tissues, i.e., adductor muscle, gill, siphon, and other nondigestive tissues.

The animal tissues were vacuum freeze-dried at − 20 °C for more than 3 days, then powdered. These samples were used for stable isotope analysis without further degreasing with organic solvent, because they already had been degreased during long-term ethanol preservation (Syväranta et al. 2008). The limnoriid isopod specimens were acidified with 1 N HCl to remove calcium carbonate from the exoskeleton, because the value of δ13C in calcium carbonate is considerably higher than that in muscle.

Once manually broken into small pieces, the wood samples were immediately dried for about a week at 80 °C. Then, were put into polyvinyl bags and kept at room temperature for 4 to 9 years. After this long-term storage, no visible mold was found from the wood samples. Before stable isotope analysis in 2016–2017, the sapwood sections where were not heavily attacked by limnoriids were sub-sampled and dried again for 2 days at 80 °C. The dried sub-samples were minced with scissors for stable isotope analysis. The seawater samples were filtered on precombusted (485 °C, 4 h) GF/F filters on the day of sampling and kept at − 20 °C until further processing in 2016. After drying for 2 days at 60 °C, the filtered POM samples were put in a desiccator for 3 h with 12 N HCl fumes to remove inorganic calcium carbonate and dried overnight at 60 °C.

Stable isotope analysis

Stable isotope analyses were conducted using a mass spectrometer IsoPrime 100 (Elementar UK Ltd., Cheadle, UK) coupled with an elemental analyzer vario MICRO cube (Elementar Analysensysteme, Lagenselbold, Germany). The δ13C and δ15N values of all animal specimens were measured using samples ranging from 0.2 to 1.2 mg dry mass. Because of the small nitrogen content, the δ13C and δ15N values of the wood logs were measured using subsamples of about 1 and 40 mg dry mass, respectively. To ensure data quality, we omitted animal samples with high C/N ratios (> 4), because these samples were suspected of being contaminated by lipids or woody particles or being altered in long-term preservation in ethanol. Considering measurement accuracy, the stable isotope ratios were rounded to the nearest decimal place.

Data analysis

Differences in values of δ13C and δ15N among bivalve species and in shell length among teredinid bivalves were statistically evaluated by one-way analysis of variance (ANOVA), and post hoc Tukey–Kramer multiple comparison tests were used to determine which pairwise comparisons of species were significant. The differences in values of δ13C among body tissues of the pholadids were tested by a paired t test.

To clarify the relative importance of potential carbon sources, i.e., wood logs and POM, for the teredinids, we applied a multiple linear regression model to the δ13C dataset. Seasonal variations in the δ13C values of POM might be reflected in the δ13C values of the teredinids after several months of delay. The full model was therefore assumed to be “Wood + POM X months before (X = 0, 2, 4, mean 0 and 2 or mean 2 and 4).” In this model, POM mean 0 and 2 months before represented the mean value of the stable isotope ratios of POM collected during the sampling month of the teredinids and 2 months before. If we needed δ13C values of uncollected POM to run the model, we replaced the value with the mean δ13C value of POM collected in the same month in other years. The best model was identified through forward stepwise variable selection based on the Akaike information criteria (AIC) procedure. After checking for multicollinearity between explanatory variables based on the variation inflation factor (VIF), we compared the effect of each potential carbon source using the standardized partial regression coefficient of the best model.

For nitrogen, the teredinids have another potential source, i.e., nitrogen fixation by symbiotic microbes. Its δ15N value is incorporated into the intercept of the model, because the δ15N value of dissolved dinitrogen in seawater is constant at around 0.5‰ in coastal areas (Naqvi et al. 1998). This makes it impossible to quantify the contribution of each potential nitrogen source to the diet of the teredinid bivalves. The δ15N dataset was therefore tested under the same procedure to check whether wood logs and/or POM contribute to their nitrogen sources.

We also applied our dataset to a widely used Bayesian stable isotope mixing model, in which the TEF values were quoted from Charles et al. (2018). We then compared the resulting proportional contributions with the values obtained in the multiple linear regression model. These statistical tests were conducted in R ver. 3.5.1 (R Core Team 2018), using packages such as “multcomp” for ANOVA and the Tukey–Kramer test and “siar” for the Bayesian stable isotope mixing model approach.

Results

δ13C and δ15N values of wood borers and non-wood-boring bivalves

The mean δ13C and δ15N values of the teredinid bivalves were lower than those of POM and higher than those of the wood logs (Fig. 1). In contrast, the non-wood-boring bivalves had higher mean δ13C and δ15N values than POM. The δ13C and δ15N values of each teredinid species were significantly different from those of non-wood-boring bivalves (Table 1) (one-way ANOVA, P < 2e-16, F = 49.76 for δ13C and 21.86 for δ15N, and Tukey–Kramer, P < 0.001), except for the relationships between T. bartschi and the two non-wood-boring bivalves, B. mutabilis and C. gigas, in terms of δ15N (Tukey–Kramer, P = 0.32 and 0.10, respectively). δ13C values of M. striata was characterized by a wide variation, including both values characteristic of the non-wood-boring bivalves and values characteristic of the teredinid bivalves (Fig. 1), and its δ15N value was positively correlated with the δ13C value (Fig. 1). The mean δ13C value of limnoriids, including both L. tuberculata and L. saseboensis, was significantly lower than that of non-wood-boring bivalves (t test, P = 1.75e-09) and slightly higher than that of teredinid species (Fig. 1). The mean C/N ratio in the limnoriids was significantly different from (t test, P = 7.73e-14) and higher than that in all bivalve species (Table 1).

Fig. 1
figure 1

Carbon and nitrogen stable isotope ratios of marine wood-boring invertebrates and non-wood-boring bivalves. Error bars: standard deviations: Teredo navalis (N = 30: closed circle), Teredo bartschi (N = 30: closed square), Teredo clappi (N = 30: closed triangle), Lyrodus pedicellatus (N = 30: closed inverted triangle), Nototeredo edax (N = 3: closed diamond), Martesia striata (N = 20: gray circle), Limnoriids (N = 5: gray square), non-wood-boring bivalves (N = 16: open triangle), Particulate organic matter (N = 26: open circle), and Cryptomeria japonica (N = 45 for carbon and 33 for nitrogen: open square). For Martesia striata, gray circles denote individuals

Full size image
Table 1 Results of stable isotope analysis of animals and potential food source
Full size table

Carbon sources for teredinid bivalves

The δ13C values did not significantly differ among the teredinid species (one-way ANOVA, F = 1.677, P = 0.16), and therefore in the following statistical analysis we tested the carbon source not for each teredinid species but for the teredinids altogether. The δ13C values of the wood logs ranged from − 28.8 to − 24.5‰ (Table 1), and there was also wide seasonal variation in the δ13C values of POM, ranging from − 21.9 to − 19.2‰. The δ13C values of both potential carbon sources were significantly and positively related to those of the teredinids (minimum − 25.2‰, maximum − 20.2‰) (P < 2.2e-16 for wood, and P < 3.614e-11 for POM, respectively) (Fig. 2). Based on AIC, the best model of the carbon source for teredinids was identified as “Wood + POM 0 month before” (Table S1), and the δ13C of the teredinid bivalves was best explained by the following multiple linear regression model formula (1), in which all explanatory variables were statistically significant (P < 0.05) (Table S1):

$$\delta ^{{13}} C_{{{\text{Teredinids}}}} = 0.69 \times \delta ^{{13}} C_{{{\text{Wood}}}} + 0.38 \times \delta ^{{13}} C_{{{\text{POM 0 month before}}}} + 3.70$$
(1)
Fig. 2
figure 2

Correlation between carbon stable isotope ratios of the teredinid bivalves (N = 123) and their potential food sources in a woods and b particulate organic matter (POM)

Full size image

The VIF value between δ13C wood and δ13C POM 0 month before was only 1.07, which resulted in rejection of the multicollinearity between them. Using partial regression coefficients (0.69 for wood logs and 0.38 for POM), we calculated the standardized partial regression coefficients. The standardized partial regression coefficient of wood logs (0.64) was larger than that of POM (0.39). To determine the contribution of POM to their carbon source, we canceled the effect caused by the variation in δ13C values of wood logs by using the following correction formula (2) based on the regression line in Fig. 2a:

$$\delta ^{{13}} C_{{{\text{Teredinids}}}}^{*} = \delta ^{{13}} C_{{{\text{Teredinids}}}} - \left( {\delta ^{{13}} C_{{{\text{Wood}}}} \times 0.802 - 1.092} \right) + {\text{mean}}\,\delta ^{{13}} C_{{Teredinids}}$$
(2)

The corrected δ13C values (hereafter referred to as δ13C*) of the teredinids were high from July (early summer) to September (late summer) and fluctuated seasonally, synchronously with the δ13C values of POM (Fig. 3). When we compared the δ13C* values among the teredinid species, that of N. edax was significantly different from (one-way ANOVA, P = 0.0089, F = 3.56, Tukey–Kramer, P < 0.05) and higher than those of the other teredinid species (Fig. 4a).

Fig. 3
figure 3

Monthly changes in carbon stable isotope ratios of the teredinid bivalves (N = 123: closed circle) and particulate organic matter (POM) (open circle). Error bars: standard deviations: δ13C* refers to corrected δ13C values. Further details are given in the text

Full size image
Fig. 4
figure 4

a Carbon and b nitrogen stable isotope ratios among the teredinid bivalves. Different letters (A–C) indicate significant differences by the Tukey–Kramer test (P < 0.05). Horizontal lines in each box plot represent third quartile, median and first quartile. The whiskers extend to 1.5 × interquartile range. Dots represent outliers

Full size image

We also investigated the size-related δ13C* values of each teredinid species. Small individuals had wide variation in δ13C* values, and their δ13C* values converged to relatively high values as they grew (Supplementary Fig. 1). Among the four teredinid bivalves except N. edax, of which sample size was small, T. bartschi [shell length (SL) 5.7 ± 1.4 mm (mean ± standard deviation)] was the largest species in shell length, and its shell length was significantly different from that of the smallest teredinid species, T. clappi (SL 4.2 ± 0.7 mm) (one-way ANOVA, P = 0.00018, F = 7.22, Tukey–Kramer, P < 0.001). The remaining two species, T. navalis and L. pedicellatus, had medium shell lengths of 4.9 ± 1.6 and 4.9 ± 1.5 mm, respectively.

Nitrogen sources for teredinid bivalves

δ15N values differed significantly among the teredinid species (one-way ANOVA, P = 5.6e-10, F = 15.07). The lowest δ15N value [5.3 ± 0.8‰ (mean ± standard deviation)] characterized T. navalis, and T. bartschi had the highest δ15N value (7.1 ± 1.1‰) (Fig. 4b). The δ15N values of T. clappi and L. pedicellatus were intermediate (6.4 ± 0.9‰ and 6.3 ± 0.9‰, respectively). The δ15N values of wood logs and POM were widely variable, ranging from − 3.3 to 5.1‰ and from 4.0 to 8.4‰, respectively (Table 1). Using these variations in potential nitrogen sources, we modeled changes in δ15N of each teredinid species except N. edax, of which sample size was small, according to the following multiple linear regression model formulas (3–7):

Teredo navalis:

$$\delta ^{{15}} N = 0.40 \times \delta ^{{15}} N_{{{\text{POM}}\,{\text{mean}}\,{\text{2}}\,{\text{and}}\,{\text{4}}\,{\text{month}}{\kern 1pt} {\text{before}}}} - 0.12 \times \delta ^{{15}} N_{{{\text{Wood}}}} + 3.11$$
(3)
$$~\delta ^{{15}} N = 0.47 \times \delta ^{{15}} N_{{{\text{POM}}\,{\text{mean}}\,{\text{2}}\,{\text{and}}\,{\text{4}}\,{\text{month}}{\kern 1pt} {\text{before}}}} + 2.44$$
(4)

Teredo bartschi:

$$~\delta ^{{15}} N = 0.58 \times \delta ^{{15}} N_{{{\text{Wood}}}} + 0.45 \times \delta ^{{15}} N_{{{\text{POM}}\,{\text{2}}\,{\text{and}}\,{\text{4}}\,{\text{month}}{\kern 1pt} {\text{before}}}} + 3.18$$
(5)

Teredo clappi:

$$~\delta ^{{15}} N = 0.70 \times \delta ^{{15}} N_{{{\text{Wood}}}} + 0.19 \times \delta ^{{15}} N_{{{\text{POM}}\,{\text{4}}\,{\text{months}}\,{\text{before}}}} + 3.11$$
(6)

Lyrodus pedicellatus:

$$~\delta ^{{15}} N = 0.41 \times \delta ^{{15}} N_{{{\text{POM}}\,{\text{mean}}\,{\text{2}}\,{\text{and}}\,{\text{4}}\,{\text{month}}{\kern 1pt} {\text{before}}}} + 3.76$$
(7)

If the consumer assimilates nitrogen compounds from the diet, its coefficient never becomes negative. The relationship between T. navalis and wood logs in formula (3), therefore, could be regarded as a false correlation, and we also show the second best-fitting model formula (4). In these multiple linear regression models, δ15N POM mean 2&4 months before was statistically significant for three teredinid species (P = 0.014 for T. navalis, P = 0.010 for T. bartschi, and P = 0.038 for L. pedicellatus) (Table S2). Among them, the δ15N value of wood logs was also selected in the best model (formula 5) and was positively correlated with that of T. bartschi (P = 9.6e-07, Table S2). For another teredinid species, T. clappi, δ15N POM 4 months before was selected as well as that of wood logs (formula 6), but only the latter was statistically significant (P = 2.3e-05) (Table S2). The intercept was always statistically significant (P < 0.01) for every teredinid species, whereas the coefficients differed among the teredinid species, ranging from 2.44 for T. navalis to 3.76 for L. pedicellatus (Table S2).

Carbon sources for pholadid bivalves

The δ13C values of the pholadids ranged from − 21.4 to − 16.7‰ (Table 1), and the δ13C values increased as the bivalves grew (Fig. 5a). The mean δ13C value of whole body of small pholadid individuals (SL < 5 mm; − 19.2 ± 1.2‰) was low and was significantly different from that of adductor muscles of large individuals (SL > 5 mm; − 17.4 ± 0.9‰) (Table 1) (t test, P = 0.015). Among the soft body tissues of large, full-grown specimens, the non-intestinal tissues had the lowest mean δ13C and δ15N values (Fig. 5b), which were significantly different from those of the adductor muscles (paired t test, P = 0.013 for δ13C, and 0.021 for δ15N). The gaps between them were only 0.7‰ for δ13C and 1.3‰ for δ15N (Table 1).

Fig. 5
figure 5

a Shell length-related variation in carbon stable isotope ratios of the pholadid bivalve, Martesia striata (N = 20). Note that whole body without intestines (gray small circle) was used for small individuals. b Carbon and nitrogen stable isotope ratios of M. striata tissues, i.e., muscle (N = 3: gray circle), gill (N = 3: gray diamond), siphon (N = 3: gray inverted triangle), and other tissues without intestine (N = 3: gray triangle). Error bars: standard deviations

Full size image

We also performed fine dissection (N = 10) and scanning electron microscopy observations (N = 4) preliminary over the alimentary system of ten preserved specimens ranging from 5.6 to 23.5 mm in shell length, with and without gonad/callum development, to obtain morphological evidence. Details will be reported separately elsewhere, but tiny but distinct reniform or sac-like cecum was found on the right wall of the stomach in smaller immature individuals (< 15 mm in shell length, N = 7), and its stomach, digestive tracts and the cecum were filled with ingested fragments of wood and filter-fed particles such as fragments of diatom frustules (Supplementary Fig. 2c, d). Wood debris were the predominant contents (Supplementary Fig. 2d, e) and were readily recognized through the tracts in having brownish color [Supplementary Fig. 2b: (gc) and (sc)]. On the other hand, the cecum was rudimentary or even absent in full-grown mature individuals that had developed callum (9.5–23.5 mm in shell length, N = 3), and there were few wood fragments in their tracts.

Discussion

POM contributes to the diet of marine wood borers in temperate shallow water

The teredinids have long been believed to be obligate xylophages, whereas a feeding experiment using 14C-labelled phytoplankton (Pechenik et al. 1979) and some anatomical studies of the alimentary systems and gill structures (Turner 1966; Lopes et al. 2000) suggested that they used POM as a part of their food resources. The present study supports the conclusion that wood-derived organic carbon is the main carbon source for the teredinids, by comparing the standardized partial regression coefficients of wood logs and POM in the best carbon source model of multiple linear regression analysis. However, seasonal synchronicity between the δ13C* value of the teredinids and the δ13C value of POM strongly supports the non-negligible contribution of POM to the carbon sources of the teredinid bivalves. A low VIF value guaranteed the independence of explanatory variables so we regard their standardized partial regression coefficients as indicators of their relative importance as carbon source for the teredinids. The contribution is calculated to be 62.1% for wood and 37.9% for POM based on the following formulas: wood contribution (%) = 0.64/(0.64 + 0.39) × 100 and POM contribution (%) = 0.39/(0.64 + 0.39) × 100. Thus, autochthonous production of POM strongly contributes to the diet of wood-boring invertebrates, helping them to decompose wood logs in temperate shallow water, particularly in its initial phase, where the teredinids are known to play an important role (Nishimoto et al. 2015).

Unlike the sedentary teredinid bivalves, free-living limnoriid isopods, serving as one of the keystone members in decomposing wood logs in the sea, do not live in one permanent shelter. Such a mode of life eventually causes the collapse of the honeycombed wood logs (Nishimoto et al. 2015). These limnoriid isopods have endogenous lignocellulase activity (King et al. 2010), and their digestive tracts are always filled only with woody particles (e.g., Sleeter et al. 1978). Therefore, they have long been believed to depend exclusively on woody debris as their carbon source. In the present study, the mean δ13C value of the limnoriids is slightly higher than that of the teredinids, in which 37.9% of carbon is derived from POM. It seems to be reasonable to conclude that the limnoriid isopods are not obligate xylophagous animals and partially depend on other food sources. For example, free-living xylophagous microorganisms, such as fungi and bacteria may also provide a supplemental nitrogen to the limnoriids. In addition, their exoskeletons were often covered with ciliates, fungal hyphae, bacteria, and microphyte remains, and the limnoriid isopods might use these organisms as carbon and nitrogen source by grooming their exoskeleton (Charles et al. 2019). These organic matters on the exoskeleton may also affect δ13C and δ15N values of the limnoriids specimens themselves. Indeed, most limnoriids samples showed high C/N ratios over 4, and then removed from this study to maintain data quality. In the present, it is technically difficult to measure the stable isotope ratios of these potential food sources due to their insufficient weight. Future studies should also avoid the contamination of these organic matters, for example, by using the limnoriid specimens immediately after molting.

Martesia striata changes its feeding strategy from xylophagous to filter feeding as it grows

Unlike the teredinids, a wood-boring pholadid bivalve, Martesia striata, has been thought to be a filter feeder judging from morphological features, such as the lack of a wood-storing cecum in the stomach (Purchon 1956; Turner and Johnson 1971). In the present study, however, the smaller young individuals of M. striata had relatively low δ13C and δ15N values. The low δ13C and δ15N values are not explained by the small variation among body tissues nor by possible contamination of the tissue samples by woody pellets with low nitrogen content, respectively.

Our preliminary anatomical observations first revealed the presence of a tiny but distinct reniform cecum in smaller immature specimens, whose contents of the cecum, stomach and digestive tracts were dominated by fragments of wood, as in the exclusive wood-eating teredinids and xylophagaids. The cecum present in juvenile Martesia appears to have been overlooked previously because of its tiny size and fragility, or because samples of small young individuals have not been considered (Purchon 1956; Turner and Johnson 1971). This finding in this study indicates that wood logs provide young M. striata with the carbon that is necessary for their initial growth. The δ13C values of large, full-grown individuals were relatively high, suggesting that the importance of POM as their food source increased as they grew. After they terminate boring activity, the callum covers their foot and they become true filter feeders as previously believed (Turner 1955; Turner and Johnson 1971).

Species-specific characteristics of feeding strategies among shipworms

Teredinid species have been thought to switch their feeding strategy from xylophagous to filter feeding under crowding or other environmental circumstances (Turner 1966; Lopes et al. 2000). This study provides another aspect of the diversity in feeding strategies in the teredinid species.

Specimens of N. edax have obviously higher δ13C* values than the other four teredinid species, indicating that N. edax depends on POM for its carbon sources more than other teredinid species. Nototeredo edax is the only free-spawning species at this study site, and it tends to occur in open waters (Scheltema 1971; MacIntosh et al. 2012). Especially in open waters, with sparse distribution of driftwood, it is advantageous for the species to increase their fecundity (Scheltema 1971; MacIntosh et al. 2014). We think it is reasonable to assume that spawning species rely more heavily on POM than brooding species, which helps them to maintain their large body size and consequent fecundity. Future studies need to investigate the anatomical relevance of interspecific differences in stable isotope approaches.

Because of the low nitrogen content of woody tissues, it has been widely believed that endosymbiotic nitrogen-fixing bacteria are the major nitrogen source for the teredinids (Gallager et al. 1981; Waterbury et al. 1983; Lechene et al. 2007). If nitrogen fixation was the only source of nitrogen for the teredinids, their δ15N values would be similar among the teredinid species. This study, however, has shown that δ15N values are significantly different among teredinid species, which suggest that nitrogen source is different between the teredinid species collected from a single location. The multiple linear regression models indicate that the supplemental nitrogen sources of four teredinid bivalves are diverse.

δ15N POM mean 2 and 4 months before values were positively correlated with δ15N values in three teredinid bivalves, suggesting that some nitrogen is assimilated from POM following a 2- to 4-month delay. Nitrogen is more valuable than carbon for xylophages, so it seems to take longer for turnover than carbon. For T. navalis and L. pedicellatus, POM mean 2 and 4 months before is the only parameter adopted in the best model, but the intercepts are considerably different between them, being high for L. pedicellatus and low for T. navalis. The low δ15N value is attributed to the high contribution of nitrogen fixation, in which δ15N value is apparently lower than that of POM (Naqvi et al. 1998). Therefore, T. navalis depend on nitrogen fixation for its nitrogen more than L. pedicellatus. The former species is the only short-term brooder collected at this study site and has pelagic larval stage for 10–15 days (Calloway and Turner 1988). Compared to long-term brooder with larval stage from a few hours to days, this species is not capable of local retention. Suppressing the relative importance of wood logs and POM as nitrogen sources, T. navalis may be able to bore into wood logs, even if the conditions in the recruitment area are somewhat unfavourable.

Unlike these two teredinid bivalves, the δ15N value of T. clappi is positively correlated only with that of wood logs. Wood logs have been ignored as the nitrogen source for xylophages; however, they serve as a supplemental nitrogen source for T. clappi, despite their low nitrogen content. Yamanaka et al. (2015) found that wood-derived nitrogen also served as the main nitrogen source in a deep-sea xylophage of the Xylophagaidae, Xyloredo teremachii (Iw. Taki & Habe, 1950). A feeding strategy that does not rely on POM as a nitrogen source probably have resulted in T. clappi having the smallest shell length among the four teredinid species, but this feeding strategy seems to be adapted to waters with low productivity. Contrary to other three teredinid species, the δ15N value of T. bartschi, the largest teredinid species at this study site, is positively correlated with that of both wood logs and POM. By using several nitrogen sources in addition to nitrogen fixation, T. bartschi appears to be the most highly adapted to this experimental condition among the four teredinids species.

This study attributes all the variation in C and N source utilization to host phylogeny without considering other sources of variability. This is because all teredinid specimens, regardless of species, were collected from a limited number of wood logs deployed at the same site, and it is unlikely that there are significant differences in environmental conditions between the teredinid species. However, it is important to bear in mind that these results may vary depending on environmental conditions. In the Xylophagaidae, a deep-water sister taxon of the teredinid bivalves, all members have the reduced labial palps and gills, indicative of the low filter-feeding function (e.g., Turner and Johnson 1971). However, their δ15N values tend to be higher in the specimens collected from shallower sites, which indicates the contribution of other nitrogen sources, such as POM (Voight et al. 2020). That is, the feeding value of each potential food source for xylophagous animals is not universal, and change depending on the environmental conditions. In this study, we estimated that 37.9% of carbon is derived from POM in temperate shallow water, but the degree of contribution also might change depending on the conditions, such as in-situ productivity and symbiotic community composition.

Sampling and pretreatment for appropriate estimation of food sources

Paalvast and van der Velde (2013) concluded that T. navalis collected from the polyhaline zone in the Netherlands was a filter feeder, based on their stable isotope analysis. The δ13C value of T. navalis (− 23.13‰) was considerably higher than that of wood logs (− 26.41‰) and was comparable to or slightly higher than those of known filter-feeding bivalves collected from the same site (− 23.46‰ for Mytilus edulis Linnaeus, 1758 and − 24.58‰ for C. gigas). In brackish waters, however, the δ13C value of POM is relatively low owing both to the high concentration of terrestrial input and to autochthonously produced phytoplankton with low δ13C values (Chanton and Lewis 2002; Fry 2002), which reduce the gaps in δ13C values between wood logs and POM. The small gaps in δ13C of potential food sources make it difficult to correctly estimate the contribution of each food source.

Various protocols are available for pretreatment of samples for stable isotope measurements (Caut et al. 2009). Some studies use only muscles, whereas others use the whole body. Additionally, some studies carried out premeasurement degreasing and/or acidification protocols. These different pretreatment protocols make it difficult to compare the stable isotope ratios among studies (e.g., Post et al. 2007). Here, we consider that the C/N ratio of target species and its standard deviation are important. Post et al. (2007) argued that C/N ratios could be lower than 3.5 when lipid content is consistently low in aquatic animals. Moreover, when there are large variations in C/N ratios among specimens, the samples are heterogeneous in chemical composition, and it becomes unclear what the δ13C and δ15N values represent. Charles et al. (2018), however, estimated the food resources in B. carinata using specimens with high C/N ratios and high SD values (8.3 ± 2.5). The δ13C value of B. carinata (− 27.5 ± 0.6‰) might be underestimated, resulting in further misinterpretation of the proportional contributions of carbon sources in the mixing model approach. Pretreatment that reduces the C/N ratio and its standard deviation in target species is therefore indispensable for comparable and repeatable studies. In the present study, we measured δ13C and δ15N values using ethanol-preserved animal muscles, and the C/N ratios were often lower than 3.5, with small SD values, except for the limnoriids.

Risk of the use of uncertain TEF values in the mixing model approach

Recent stable isotope mixing model approaches in Bayesian frameworks, such as the SIAR algorithm (Parnell et al. 2010), can yield probabilistic estimates for the proportional contributions of each food source to the target species, even if the number of potential food sources exceeds n + 1 (where n is the number of isotopes employed). These approaches are widely accepted for estimation of animal food sources. However, they are often based on many assumptions in setting the parameters. Charles et al. (2018) reported that a fully marine teredinid, B. carinata, derived 26% of its carbon from POM, based on a Bayesian stable isotope mixing model. In their model, the δ13C and δ15N values of both potential food sources (i.e., wood logs and POM) were used as end members, and the TEF values were quoted from previous studies dealing with the whole bodies of invertebrates, i.e., 0.30 ± 0.21‰ for δ13C and 2.5 ± 0.25‰ for δ15N (Caut et al. 2009). Concerning the assimilation of dissolved N2 by bivalves with the help of symbiotic bacteria, they assumed that the TEF was − 1‰ between dissolved N2 and fixed N and there was no fractionation between the endosymbionts and their host. When we carried out the Bayesian stable isotope mixing model using our dataset under the same settings, POM accounted for 59.2% of the teredinids carbon source (Supplementary Figs. 3 and 4). This ratio is considerably different from that of B. carinata in Charles et al. (2018) and the output from the multiple linear regression model approach in this study. These facts clearly indicate that the TEF values used in Charles et al. (2018) are not applicable for estimating the food resources in marine xylophagous animals. Thus, the mixing model should be used with care only when the TEF values and stable isotope ratios of the end members are correctly assigned (Ben-David and Schell 2001). Contrary to the mixing model approach, for carbon source analysis in the teredinids bivalves, the multiple linear regression model approach is advantageous in that it can estimate the proportional contribution of each potential carbon sources without use of the uncertain TEF values.

Speculations

This study highlights that marine wood borers depend on autochthonous production as their source of carbon and/or nitrogen in a coast of Tanabe Bay, central Japan. We assume that the same setting occurs in temperate shallow waters around the world, and offer the novel hypothesis that the transfer of organic carbon from coarse woody materials into marine ecosystems is controlled by the productivity of the ocean. This hypothesis also implies that the importance of wood logs as a food source increases in low-productivity conditions such as the deep-sea floor. By comparing wood-boring activities and autochthonous productivity among sites, we can elucidate the ecological role of terrestrial input in the form of coarse plant debris into marine ecosystems.